Trends on Application Performance Monitoring
In the era of cloud computing, big data, scalable applications and Internet of Things (IoT), it is pretty clear that software has a key role on the businesses' performance, it doesn't matter if it is a startup or an Enterprise. As the famous quote says: "Software is eating the world". Consequently, the performance of the systems is becoming increasingly relevant.
As an example, tests performed by Amazon revealed that the increase of 100 milliseconds in your e-commerce site loading time generated a 1% loss in sales. Every millisecond counts.
Along with all these technological developments came new challenges and needs, which forced the evolution of the application performance management tools (APM).
For a long time, server monitoring tools formed the basis for monitoring the IT infrastructure efficiency in a company. Those tools capture metrics such as CPU usage, disk, memory and network, allowing to identify which resources are being used by which application or service.
Solutions focused solely on the performance of the servers no longer meet current needs. What matters now is the performance of the application and the current tools need to be able to identify if there is a bottleneck that compromise their performance.
“Facebook has over one million servers. If one goes down, it doesn’t matter. What matters is how the app that is being served is performing. Is there a bottleneck or is it doing okay?” says Mr. Horowitz, Andreessen Horowitz partner.
This new paradigm has given rise to a generation of APM tools that focus exactly on application performance, identifying its major bottlenecks. These tools allow you to identify the current conditions of the application's performance, revealing if everything is working well or if there are any performance issues.
This paradigm shift is only the beginning of a long journey in the evolution of the tools of this type of monitoring. A recent survey conducted by TRAC Research revealed that there is still a gap to be filled. The main problems highlighted in the survey were:
A large number of performance incidents is perceived by the end user before the operations team becomes aware of such a problem;
Solve performance problems still demand long effort and time;
The usability of the information presented by APMs is still deficient.
Most of the tools available in the market use a descriptive approach and present a lot of information in their dashboards, many of which are irrelevant to problems' resolution.
The path to be followed by APMs is through a predictive approach, which aims to present information that can be turned into action to solve problems proactively. Using machine learning techniques and big data this new approach allows us to analyze large volumes of data and correlate multiple metrics, identifying behavior patterns that allow predicting potential problems.
This technology is the same used by credit card companies to identify fraud based on the usage pattern of each customer or by Facebook to offer personalized ads guided by user browsing history on the internet and by financial institutions in credit analysis to customers.
The main benefit of a predictive analysis is to enable the IT support team on time about the existence of a potential problem so they can take the necessary actions and that application's performance will not be compromised.
Soon performance management tools will not only identify problems in advance but they will be able to recommend solutions. A prescriptive approach. The recommended actions are then performed by the IT team in order to prevent performance problems.
The evolution will then move toward the self-adaptive systems. Such systems will be able to change their own behavior in response to their perception of the environment and the system itself. This tools will understand the problems and before they cause any impact on the application performance they will take the necessary actions to avoid performance incidents without human intervention. We will talk more about this trend on the future.
